How to Get Cited by Claude AI (2026)
Last updated: October 1, 2026
Short answer: Claude cites pages it can find in Brave Search (its web-search backend), reach through an open Claude-SearchBot crawler path, and extract clean answers from. To get cited, get indexed in Brave, allow Claude’s crawlers, lead every page with a direct answer, attribute claims to named sources with dates, and keep your brand facts consistent across the web. There is no paid placement.
Claude’s web search does not work like Google. It pulls from a different index, rewards a different page shape, and never sells its way in. If you want your brand named in Claude’s answers, you optimize for how Claude actually retrieves and cites, not for classic blue-link SEO. Below is the verified mechanism and the exact checklist I use with clients.
This guide is a spoke in our complete guide to Answer Engine Optimization (AEO). For the engagement that executes it, see Rank on AI.
How does Claude decide what to cite?
Answer: When a question needs current information, Claude writes its own search queries, sends them to Brave Search’s index, reads the returned pages, and writes an answer with inline citations. Your page can only be cited if it is in Brave’s index, reachable by Claude’s crawler, and structured so a specific passage answers the query.
The retrieval path is public record. Anthropic added Brave Search to its subprocessor list on March 19, 2025, and switched on web search in Claude the next day. According to Simon Willison (simonwillison.net, March 21, 2025), Claude’s web_search tool schema contains a BraveSearchParams object, and a test query returned ten results that matched Brave exactly (source).
The citation overlap is measurable. Profound’s analysis found roughly 86.7% overlap between Claude’s citations and Brave’s top results; a larger follow-up across ~35,000 citations over 400 queries found 79.2% of Claude’s cited URLs sit in Brave’s top 10, versus about 34% for Google’s top 10. The takeaway: your Google rank barely predicts Claude citation. Your Brave rank does.
Three facts follow directly:
- If you are not in Brave’s index, you cannot be cited by Claude’s live search.
- If Claude’s crawler is blocked, you drop out of the eligible set.
- If the answer is buried, Claude pulls a competitor’s cleaner passage instead.
How do you get into Brave’s index?
Answer: Submit your sitemap in Brave Search Webmaster Tools and earn real visits. Brave’s Web Discovery Project only makes a URL eligible for indexing after it is seen by at least 20 unique Brave browser users with data-sharing enabled. So Brave indexing rewards genuine traffic and clean technical access, not just submission.
Brave builds its index partly from anonymized signals contributed by real people using the Brave browser, documented in the Web Discovery Project (merj.com, “How Brave Search discovers new pages”). A page needs organic human visits before it enters the index, which is why low-traffic pages stay invisible to Claude even when they rank on Google.
Do this, in order:
- Create and verify a property in Brave Search Webmaster Tools and submit
sitemap.xml. Coverage typically fills in over 2 to 4 weeks. - Fix crawl blockers: no
noindexon money pages, fast server response, valid canonical tags. - Drive real visits to the pages you want cited (email, social, your own browser habits on Brave).
- Confirm the page actually ranks in Brave for your target query. If it is not in Brave’s top 10, treat that as the first metric to move.
Do you need to allow Claude’s crawlers in robots.txt?
Answer: Yes for search visibility. Anthropic runs three separate bots: ClaudeBot (training), Claude-User (fetches a page during a live chat), and Claude-SearchBot (indexes content for Claude’s web search). To be eligible for citation you must allow Claude-SearchBot and Claude-User in robots.txt. Each is controlled independently.
Per Anthropic’s own crawler documentation, all three bots honor robots.txt (Anthropic support). A common, costly mistake: teams block ClaudeBot to opt out of training, then accidentally block Claude-SearchBot too, removing themselves from search citations. Decide training and search separately.
A visibility-first robots.txt looks like this:
User-agent: Claude-SearchBotthenAllow: /(eligible for Claude web search).User-agent: Claude-UserthenAllow: /(Claude can fetch your page when a user asks).- Optionally
User-agent: ClaudeBotthenDisallow: /if you want to opt out of training while staying citable.
How should you structure a page so Claude extracts the answer?
Answer: Lead with the answer. Put a 40 to 75 word direct response right under each question-shaped H2, then support it with lists and tables. Claude favors reference-style pages with descriptive headings and concise, self-contained passages it can lift cleanly. Bury the answer under setup and Claude quotes a competitor instead.
Claude’s web search reads pages and extracts specific passages. The structure that wins:
- Question-shaped H2s. Match the literal phrasing people type (“How do you get cited by Claude?”), not clever headings.
- Answer-first capsule. One tight paragraph that stands alone without the rest of the section.
- Lists and tables. Steps, criteria, and comparisons in extractable form, not prose walls.
- One idea per passage. Each paragraph should make sense quoted out of context.
- Visible last-updated date near the top, so Claude can see the page is current.
What trust signals make Claude pick your page?
Answer: Named authors with real credentials, original or primary data, inline “According to [source, date]” attributions, visible publish and update dates, and brand facts that match across your site, Wikipedia, LinkedIn, and directories. Claude leans on reference-grade, verifiable content, so the more your page reads like a citable reference, the more it gets cited.
These are the E-E-A-T signals that carry into AI answers:
- Byline a real expert with a credential line and a linked bio. Anonymous content is weaker to cite.
- Publish original data. A number only your page reports is the single strongest citation magnet, because Claude has nowhere else to source it.
- Attribute every external claim inline: “According to [named source, date], [fact]” with a link. This is exactly what Claude does, and it mirrors pages built the same way.
- Keep brand facts consistent (name, founding, offering, location) across your site, Wikipedia/Wikidata where eligible, LinkedIn, and industry directories. Conflicting facts lower confidence.
- Add Article and FAQPage schema so machines parse author, date, and Q&A structure without guessing.
The Claude citation checklist
Answer: Work the table top to bottom. Index and crawler access are prerequisites (no access, no citation). Structure and trust signals decide whether Claude picks you over a competitor who is also indexed. Review it per page, not once per site.
| Lever | Action | Why Claude rewards it | Priority |
|---|---|---|---|
| Brave index | Submit sitemap in Brave Webmaster Tools; earn real visits | Brave is Claude’s web-search backend; ~79% of cited URLs are in Brave’s top 10 | Prerequisite |
| Crawler access | Allow Claude-SearchBot and Claude-User in robots.txt | Blocked bots cannot index or fetch your page | Prerequisite |
| Answer-first | 40 to 75 word direct answer under each question H2 | Claude extracts self-contained passages | High |
| Question headings | Phrase H2s as the queries people type | Matches the retrieval query literally | High |
| Original data | Publish a stat, study, or benchmark only you report | No alternative source means forced citation | High |
| Named author | Expert byline + credential + linked bio | Author authority is a trust signal | High |
| Inline citations | “According to [source, date]” with links | Mirrors Claude’s own sourcing pattern | Medium |
| Freshness | Visible last-updated date; refresh content | Current pages beat stale ones | Medium |
| Structured data | Article + FAQPage JSON-LD | Machine-parseable author, date, Q&A | Medium |
| Brand consistency | Match facts across site, Wikipedia, LinkedIn, directories | Conflicting facts lower confidence | Medium |
What does not get you cited by Claude?
Answer: You cannot pay for placement; Anthropic sells no ad slots or sponsored citations in Claude. Classic link-building for Google rank, keyword stuffing, and thin “best X” listicles do little, because Claude pulls from Brave and prefers reference-grade passages. Chasing Google tactics while ignoring Brave is the most common wasted effort.
Skip these:
- Paying for a citation. There is no paid placement. Any vendor offering guaranteed Claude citations for a fee is selling something Anthropic does not.
- Google-only SEO. Google top-10 position predicts Claude citation far more weakly than Brave position does.
- Thin aggregator pages. Generic “top 10” posts with no original substance rarely get lifted.
- Blocking the wrong bot. Opting out of training is fine; accidentally blocking
Claude-SearchBotremoves you from search.
Frequently asked questions
What search engine does Claude use for web search?
Claude’s web search runs primarily on Brave Search, added to Anthropic’s subprocessor list on March 19, 2025. A second subprocessor, TurboPuffer, handles vector and full-text retrieval. Your Brave Search ranking is the strongest predictor of whether Claude cites you.
Can I pay Anthropic to be cited by Claude?
No. Anthropic offers no paid placement, sponsored citations, or ad slots in Claude’s answers. Citation is earned through indexing, crawler access, and citable content. Any “guaranteed Claude citation” offer is not a real Anthropic product.
Which robots.txt bot controls Claude citations?
Claude-SearchBot indexes content for Claude’s web search, and Claude-User fetches pages during live chats. Allow both to stay citable. ClaudeBot is the training crawler and can be blocked separately without losing search visibility.
How long until Claude can cite a new page?
Plan on 2 to 4 weeks after you submit your sitemap to Brave Search Webmaster Tools and the page earns real visits. Brave’s Web Discovery Project needs genuine human traffic before a URL becomes eligible, so brand-new, zero-traffic pages take longest.
Does ranking on Google mean Claude will cite me?
Not reliably. In one ~35,000-citation study, about 79% of Claude’s cited URLs were in Brave’s top 10 versus roughly 34% in Google’s top 10. Optimize for Brave and for citable page structure, not Google rank alone.
What page structure does Claude prefer to cite?
Reference-style pages with question-shaped H2s, a short direct answer under each heading, and extractable lists and tables. Claude lifts self-contained passages, so every answer should make sense quoted on its own, with a visible update date and named author.
Does original data really increase Claude citations?
Yes. A statistic, survey, or benchmark that only your page reports gives Claude no alternative source, which makes citation close to forced when the query matches. Original data is the highest-leverage single asset for AI-search visibility.
How do I know if Claude is already citing my brand?
Ask Claude the questions your buyers ask and check which sources it names, then repeat the queries in Brave Search to see your ranking. A professional AEO audit tracks citation share across Claude, ChatGPT, and Google AI over time.
Want your brand named in Claude’s answers? This page is built to the exact standard it describes, because that is the service. See how we run the full program on Rank on AI, read the complete AEO guide, compare the playbook for getting cited by ChatGPT, or book a consultation to map your citation gaps.
Sources
- Simon Willison, “Anthropic’s use of Brave Search,” simonwillison.net, March 21, 2025, link
- Anthropic, crawler documentation (ClaudeBot, Claude-User, Claude-SearchBot and robots.txt), link
- Merj, “How Brave Search discovers new pages: the Web Discovery Project”, link
- Profound citation-overlap analyses of Claude vs Brave and Google results (86.7% overlap; 79.2% in Brave top 10).
- Xponent21, “Anthropic lists two web-search subprocessors: Brave and TurboPuffer”, link